Search papers, labs, and topics across Lattice.
4
0
9
Speculative decoding's speed boost just got a whole lot bigger: DIVERSED dynamically loosens the verification constraints, letting more good tokens through and accelerating inference.
Current LLM agent safety benchmarks are missing over 20% of unsafe behaviors, even after agents pass the benchmark.
Ditch slow, verbose language-based action planning: this diffusion model directly predicts discrete, kinematically feasible waypoints for faster, more precise autonomous driving.
Two-stream attention in any-order autoregressive models isn't just about separating position from content, but about resolving a deeper structural-semantic tradeoff that single-stream models can't handle.
